Bionic Electronic Nose Based on MOS Sensors Array and Machine Learning Algorithms Used for Wine Properties Detection.

Area of Science:

• Analytical Chemistry
• Sensor Technology
• Machine Learning

Background:

• Wine authentication and quality control are crucial in the food industry.
• Distinguishing wines by origin, age, and grape varietal relies on complex chemical profiles.
• Objective and rapid analytical methods are needed for wine characterization.

Purpose of the Study:

• To develop a portable electronic nose (E-nose) prototype for wine odor analysis.
• To evaluate the performance of machine learning algorithms in classifying wines based on sensory data.
• To identify key wine properties detectable by the E-nose system.

Main Methods:

• Fabrication of a portable E-nose prototype utilizing metal oxide semiconductor (MOS) sensors.
• Application of four machine learning algorithms: extreme gradient boosting (XGBoost), random forest (RF), support vector machine (SVM), and backpropagation neural network (BPNN).
• Training and testing classification models for wine identification based on production area, vintage, fermentation process, and varietal.

Main Results:

• The E-nose system demonstrated effectiveness in detecting and differentiating wine odors.
• Backpropagation neural network (BPNN) achieved high accuracy (94%) for production area and (92.5%) for varietal identification.
• Support vector machine (SVM) showed the best performance for vintage (67.3%) and fermentation process (60.5%) classification.

Conclusions:

• The developed portable E-nose is a viable tool for objective wine classification.
• The choice of machine learning algorithm significantly impacts classification accuracy for different wine properties.
• Further optimization of algorithms can enhance the E-nose's capability in wine authentication and quality assessment.